Process Units River Evaluation (PURE) Method for River Corridor Delineation

Transcription

1 University of Massachusetts Amherst Amherst Water Reports Water Resources and Extension 2019 Process Units River Evaluation (PURE) Method for River Corridor Delineation John D. Gartner Christine E. Hatch Follow this and additional works at: Part of the Databases and Information Systems Commons, Environmental Health and Protection Commons, Geomorphology Commons, Hydrology Commons, Natural Resources Management and Policy Commons, Numerical Analysis and Scientific Computing Commons, and the Water Resource Management Commons

2 Process Units River Evaluation (PURE) Method for River Corridor Delineation. John D. Gartner and Christine E. Hatch University of Massachusetts Amherst March 19, 2019 DRAFT

3 Abstract River corridors are defined as areas adjacent to rivers that are likely to affect and be affected by river and floodplain processes. Geomorphic and hydrologic models can be used to delineate river corridors to address fundamental questions, such as which parts of the fluvial landscape are most likely to respond to climate and land-use change. Land managers are also increasingly using river corridors for applied purposes such as hazard avoidance, ecological conservation, and water quality protection. Currently, the most-used river corridor delineation methods rely on historic maps, field surveys, and/or calibrated empirical models which may be time-intensive, subjective, or require expert user input. Here we present a novel approach for delineating river corridors that utilizes widely available geospatial data (and thus can be applied uniformly across broad spatial extents) to predict areas susceptible to flooding and geomorphic change, which also includes critical areas for riparian ecology and water quality. Our approach is called the Process Units River Evaluation (PURE) River Corridor Delineation Method, and it is derived from the sum of five functional process units that together capture the river corridor: (i) the Flood Processes Unit, derived from hydraulic modeling to determine areas subject to overbank deposition and erosion, in-channel deposition and erosion, bank erosion, and channel avulsions; (ii) the Landslide and Steep Terrain Processes Unit, based on terrain slope to show locations subject to sediment delivery, bank failures, and other mass wasting proximal to the flood-prone area; (iii) Wetland Processes Unit, based on the U.S. National Wetlands Inventory to show areas where wetland processes occur; (iv) Channel Migration Processes Unit, based on channel location and migration rates to show areas susceptible to lateral channel movement; and (v) Riparian Ecologic Processes Unit. We examine this delineation approach in three river reaches in mountainous and low-relief watersheds in the northeastern U.S. and compare results with other river corridor and flood area delineation techniques. Predicted PURE river corridors are also tested against areas of recent geomorphic change, observed in the field and in historic imagery, and the PURE corridor performs very well, capturing 92% of observed landslide areas, 87% of observed floodplain deposition areas, and 100% of channel migration areas. Our river corridors also show general agreement with more time-intensive delineation approaches, and the resulting corridors are typically broader than areas of recent geomorphic change. The case studies indicate the PURE Corridor method is successful at providing both an accurate assessment of potential active hazard areas and sensitive environmental areas, and that it also includes a margin of safety that many managers desire. This work exhibits the potential for rapid, uniform, and objective river corridor delineation across large areas with transparent communication of the data inputs. Its modular nature allows for flexible weighting of different metrics to suit specific applications, and piecewise updating as new approaches become available.

4 1. Introduction Rivers and the swath of land next to rivers are critically important for riparian ecosystems, water resources, human infrastructure, and prone to natural hazards. Several hydrologic and geomorphic processes including flooding, erosion, deposition, river migration, bank failures, and landslides create dynamic areas where water, sediment, nutrients, carbon, and pollutants are transported, sometimes gradually and sometimes catastrophically (Bierman et al., 2014). The dynamic processes create prized ecological areas with high biodiversity, spatiotemporal patch dynamics, a variety of riparian habitats, and an ecotone from aquatic to terrestrial communities (Naiman et al., 1993; Pickett and White, 1985). This is also a region of heightened connectivity, due to the migration and dispersal of aquatic and riparian species (McCormick et al., 1998; Strayer, 2008). Physical processes also enhance connectivity, with hydrologic and geomorphic processes transporting materials upstream, downstream, and laterally between channels, floodplains, and hillslopes (Allen, 2008; Bracken and Croke, 2007; Croke et al., 2013; Gartner, 2015; Pringle, 2001; Stanford and Ward, 1993; Wohl and Beckman, 2014). The increased connectivity within the river corridor takes on a human dimension when pollutant discharge, bank alteration, dams, and other human activities in one area affect conditions elsewhere in the river corridor. Many human structures including transportation networks, farms, and homes are situated in this active river corridor area in order to take advantage of fertile soils, natural, relatively flat pathways through rugged terrain, and access to water and hydraulic power. As a result, this region is prone to natural hazards, in which floods, landslides, bank erosion, and sedimentation can damage roads, homes, farms and other infrastructure (Dethier et al., 2016; Gartner et al., 2015; Magilligan et al., 2015). These hazards existed in historic climatic conditions, and in the future, increasing precipitation and storminess may increase the frequency and magnitude of these hazards (Huang et al., 2017; Yellen et al., 2016). Indeed, the hydrologic and geomorphic effects of floods are consistently among the most costly and lethal natural hazards (Ward et al., 2017). Overall, humans benefit from, alter, and are threatened by processes in the river corridor. The term river corridor has increasingly been used to describe the dynamic and important swath of land occupied by and adjacent to rivers (Kline and Cahoon, 2010; Warner et al., 2018). Historically, there has been no single term to describe this area, as it is greater than just the channel, banks, and floodplain. The Oxford English Dictionary defines a river corridor as A narrow stretch of land comprising a river and the areas adjacent to it, especially one important as a route for movements and communications; a (narrow) river valley. Origin 1920s. This definition overlooks the geomorphic and ecological activity inherent to this area. Other terms are roughly synonymous with the term river corridor and allude to its dynamic nature. These terms include freedom space for rivers, (Biron et al., 2014; Buffin-Belanger et al., 2015), fluvial territory (Ollero, 2010), erodible corridor (Piegay et al., 2005), active river area (Smith et al., 2008), and channel migration zone (Rapp and Abbe, 2003a). Other widely used terms are not equivalent to the river corridor, even though they describe areas with special protections next to rivers. Often these areas overlap with the river corridor but are too narrow and do not consider the suite of river corridor processes. Examples include riparian area or riparian zone, from the Latin ripa, meaning banks is typically very narrowly defined as the interface the banks of a water body, the Ordinary High Water Mark (OHWM) from the U.S. Army Corps of Engineers used in the application of the Clean Water Act (Lichvar and McColley, 2008) and the Special Flood Hazard Area (SPHA) from the Federal Emergency Management Agency (FEMA) used for Flood Insurance Rate Maps. FEMA flood zones can be especially problematic as a surrogate for river corridors, because of they tend

5 focus on the likely extent of high waters during present conditions (or conditions at the time of mapping) without consideration for hazards generated by the movement and migration of river channels, including avulsing (jumping to a different channel location), bank failures or landslides, or widening as a result of landscape/ land use change, climate change, or ongoing natural processes. In an effort to develop a consistent, statewide methodology for evaluating and managing rivers in Massachusetts, in 2012 the University of Massachusetts Amherst and the RiverSmart Communities project assembled a Fluvial Geomorphology Task Force (FGM Task Force). The FGM Task Force is comprised of a University of Massachusetts Amherst steering committee and ~30 river specialists representing academia, consulting, regulatory and conservation agencies, and government at local, state and federal levels from Massachusetts and neighboring states. During three workshops between 2014 and 2017, the FGM Task Force synthesized the many ways river corridors are defined, delineated and utilized. Among the Task Force s outputs were an evaluation of river corridor delineation strategies, a roadmap for what an optimal strategy should contain, and the creation of a synoptic definition for river corridors: The river corridor is the area where channel-driven fluvial processes, including the river s natural movement of water, sediment, debris and other materials, affect or are likely to affect the landscape, based on current, historic, and projected conditions. The primary challenge when delineating river corridors is to predict future locations of river-related processes. Direct observations of river processes are difficult because they may be subtle, such as the sporadic and intermittent buildup of sediment and nutrients on floodplains. Many geomorphic processes occur only rarely, such as landslides during major storms. In addition, the probability (equivalently, the likelihood over time) that any process will occur at a given moment changes with distance from the current river channel. For example, river migration may be very likely to occur a few meters from an existing channel over the next decade, yet there is an unlikely but non-zero chance that a river may migrate hundreds or thousands of meters over century or millennial time-scales (Buraas et al., 2014; Hickin and Nanson, 1984; Schwenk et al., 2017). And, to further complicate matters, the next prediction of that same process depends heavily on exactly where the previous prediction placed the active river channel. Thus, river corridor delineations depend inherently on the time and length scales of interest. Most river corridor delineations consider human-relevant timescales, approximately annual to century time frames, but this is rarely stated explicitly. The most often-used river corridor delineation techniques utilize insights into river dynamics to predict areas where river-related processes will occur; and with these insights come some common challenges. Three common limitations associated with river corridor mapping include: scaling issues, latent subjectivity, and filtering effects. resulting from using a single metric or algorithm. Scaling issues refer to the fact that, for any single method, it is very difficult to achieve the same level of accuracy when moving from a very fine scale (e.g x river width) to a very broad scale (e.g. statewide or regional). Maps generated at one scale may miss important features at another; or be so general as to provide little insight or predictive capacity for decision-making. Latent subjectivity is the term we apply to methods that require an expert to make judgements about mapping choices and corridor placement. These methods are not able to be widely transferrable, because they depend on a few individuals who are very specifically trained to apply the technique correctly, and could potentially yield very different results in the hands of a non-expert. Filtering effects are the result of using a single metric or algorithm

6 across an entire landscape regardless of its relative complexity or variability. These techniques have input parameters that apply to an entire area of interest, and for a consistent result, are likely to be pegged to a single value. If this threshold value is too small, it may result in underrepresentation of the river corridor; if the threshold is too large, it may cover an impractically large area. One method of river corridor delineation, referred to as the Meander Migration Model, or Morphodynamic evolution model, builds on the concept that the past is the key to the future, and employs detailed analysis of historic maps, image time-series, soils, and field characteristics (Larsen, 2007; Pasquale et al., 2011; Rapp and Abbe, 2003b) to determine where river-related processes have occurred and might occur in the future. With this technique, different river corridor processes can be considered individually when experts evaluate different lines of evidence. The multiple lines of field evidence avoid the filtering effect of a single algorithm. This method has been applied successfully to specific reaches and sometimes rivers 10 s of km long, but is too time consuming to be scaled up to broad regions. In addition, latent subjectivity, scaling issues and reproducibility are challenges associated with this technique. Different experts may delineate different boundaries for the river corridor, and it is not clear how different factors are weighted by experts when determining the corridor boundary. A second method, the Vermont Stream Geomorphic Assessment Tool (SGAT) River Corridor protocol, addresses the scaling issue by specifically defining the relationship between the river corridor width and the meander belt width, which, in turn, is a function of channel width in essence larger rivers have wider corridors (Kline and Cahoon, 2010; Kline and Dolan, 2008). The Vermont River Corridor is roughly equal to the meander belt width, and defined as a swath of 2 to 8 channel widths centered on the meander centerline, with some adjustments for local conditions. For streams with drainage areas less than 2 mi 2, a standard 50-foot buffer is applied. In the Vermont method, many river corridor processes are filtered through one key algorithm the meander belt equals x channel widths from the meander centerline. This allows for automation and consistent application over large areas, but the empirical algorithm is not explicitly tied to the suite of physical river corridor processes it is predicting. The result of the algorithm may be too wide in some areas, especially in in confined valleys, and too narrow in others. In Vermont, this effect is ameliorated successfully by experts who evaluate and digitize the meander centerlines and digitize adjustments for valley confinement and human obstructions along each river. In addition, a publicly accessible database allows any landowner to propose revisions to the maps based on more accurate local conditions. Although the Vermont method has been very effective across the state of Vermont, the hand digitizing can be time consuming to scale up to very large regions. The method also has some latent subjectivity, as only a small group of trained experts perform the mapping, and even among them, different experts may digitize meander centerlines slightly differently or make different adjustments for confinements to the river corridor. A third method, the Active River Area (ARA) approach developed for the Nature Conservancy, builds from the insight that the river corridor is a function of both the distance from and elevation above the river channel (Smith et al., 2008). The parameters of distance-elevation algorithms are determined at selected locations (training sites) where the width of the river corridor has been evaluated by experts. Next, the algorithm is applied on a 30-m grid using Geographic Information Systems (GIS) across broad regions with physiography similar to applicable training sites. With this strategy, the Active River Area approach has been applied to the entire eastern seaboard of the U.S.. Although this method solved the scaling issue to cover very large regions, the 30-m grid size makes it difficult to scale down to reach-scale

7 corridor delineation for useful parcel- or town-scale application. The method also has a filtering effect: all of the river corridor processes are evaluated though a single metric, the distance-elevation algorithm. This results in the corridor being too wide in most areas and too narrow in a few others. This method also depends somewhat on the different experts choose to delineate at the training sites: their choices of very wide or very narrow corridors would then be applied across the whole region of analysis. The method is reproducible if the parameters of the distance-elevation algorithm are known and reasonable for the area of interest, but the initial step of developing the distance-elevation algorithm is less reproducible. The ARA method is particularly skilled at capturing the entire valley bottom for river systems, which, even if not exactly the river corridor, is a very useful metric for river system evaluation. We stress that each of the methods described above historical meander migration observations, Vermont SGAT, and TNC ARA show remarkable insights and advances in river corridor delineation, yet challenges still exist in scaling, subjectivity, reproducibility, and determining metrics that faithfully represent a wide range of river corridor processes. A fourth method is the most straightforward of all. The Massachusetts River Protection Act applies an algorithm protecting an area including a 200-foot buffer from each side of the river channel, except in specific urban areas where the protected area is a 25-foot or 50-foot buffer from each side of the channel. This approach benefits from being both simple and empirical, and is very easy to implement. The MA RPA is based on the general observation that most river corridor processes occur adjacent to rivers. The method is reproducible and can be scaled up efficiently to very large regions. However, the choice of 200-foot or 50-foot buffers are purely subjective. The rule is also blunt. Filtering through this one algorithm (the corridor equals a buffer x ft from the channel) tends to over-predict or under-predict the extent of river corridor processes from one location to the next. The effect could be reduced by developing a group of more physically-based rules that are more directly tied river corridor processes. Here we present a new scientific approach, based as explicitly as possible on active river processes: the Process Units River Evaluation (PURE) River Corridor Delineation Method. In developing this methodology, we aim to include measurable metrics of physical river processes and develop a protocol that could be largely automated while still applicable at both reach and landscape scales. Conceptual Map of the Process Units River Evaluation (PURE) River Corridor Delineation Method 1 List key river processes 2 Group into process units 3 Develop Mapping Rules Overlay Mapped Units Refine & 4 Sum for 5 Repeat PURE Corridor The method is comprised of five steps: 1) list key river processes, 2) group similar river processes into process units, 3) develop mapping rules for each process unit and map each, 4) overlay mapped processes units and sum them to delineate the maximum extent of the river corridor, and 5) refine as desired for the scale of use by using different data or algorithms for specific process units. This study

8 was sponsored by the US Fish and Wildlife Service (USFWS) and North Atlantic Landscape Conservation Cooperative (NALCC) to investigate river corridor delineation that can be performed consistently across the entire NALCC region across the Northeastern U.S.. The NALCC region spans 13 states along the Atlantic coast from Virginia to Maine plus West Virginia and Vermont. Although the method is developed with the NALCC region in mind, it is our intention that this method could be applied across the U.S. and beyond. In this work, we address the central questions that arised while developing this method, including: What processes should be considered for the river corridor? Which mapping rules and data can be used to predict where these processes occur? How does PURE River Corridor Delineation compare with other river corridor delineations? How does the PURE River Corridor Delineation compare with locations of known geomorphic change? 2. PURE Corridor Delineation Method The PURE Corridor Delineation method is comprised of five steps to determine areas where natural river processes are likely to occur, area that are important for riparian species, and areas that are susceptible to river-related hazards. The first two steps set up the theoretical basis and guiding principles that drive the steps that follow. These first steps are best done in a large group setting with representatives from all potential stakeholder groups, including those with competing interests and diverse opinions. Steps 3 and 4 are automated computational steps that proceed according to a computer algorithm in GIS using threshold values set in Step 3. These thresholds are chosen to be most effective at capturing the desired processes identified in Step 2. Finally, Step 5 is an opportunity to refine threshold values to improve results after comparing to other data, or to replace an initial algorithm with an alternate one should new data or more advanced methods become available. Step 1: Research, identify, and list the river-related processes that should be considered in river corridor delineation. Step 2: Group these into self-similar Process Units that capture these river processes and functions. Step 3: Create a set of rules that allow for digital mapping for each Process Unit, to result in polygons which show the region adjacent to the river where processes included in each process unit are likely to occur. Step 4: Create the total river corridor as the sum of the process unit polygons, which shows the predicted extent of all the processes listed in Step 1. Step 5: Refine the process units and corridor delineation as necessary- either for a specific desired use, or to incorporate a new advance in one or more of the Process Unit mapping rules. The Process Units approach requires that each step be completed in turn, but there is leeway within each of the steps at key decision points. For example, one might add or subtract processes in Step 1 to suit the needs of a particular research question or particular natural resource management concern. In

9 addition, one might choose one of a number of different rules that are available to map any given process in Step 3. Finally, the refinements in Step 5 can take several forms. 2.1 Step 1 Identify River Processes A central question when developing and employing any river corridor delineation method is what processes should be considered? This may vary for different users. Table 1 lists the processes that we considered as occurring within the river corridor, and which were identified as important for inclusion by the Fluvial Geomorphic Task Force in Massachusetts during several meetings. Figure 1 shows a theoretical landscape with two cross-sectional views of two areas in a river valley containing different river processes that we consider when delineating our river corridor. Many of the river processes in Figure 1 are responsible for the transport of water and sediment. These processes also remove, transport, and store pollutants, nutrients, carbon, and other materials which are mobilized with water and sediment movement. The locations of these processes are also likely to be impacted by the removal, transport, and accumulation of pollutants, nutrients, and carbon, because these materials are also mobilized with water and sediment movement (Landis et al., 2012). We briefly define each of our river processes, which correspond to the numbers in Figure 1, as: 1. Flooding is the process where high waters overtop the channel banks and flow out onto floodplains and other near-channel areas, often with enough force to mobilize sediment and other materials. 2. River migration occurs as the combination of bank erosion and point bar deposition over time, such that the river channel moves laterally. 3. Wetland processes occur where frequently saturated soils accelerate biologic processes including the decay of organic matter, the release of sulfur, nitrogen and carbon into the atmosphere, and the removal of nutrients, organic matter, and pollutants from moving water (Novitski et al., 1996). In addition to the ecological benefits, wetlands are low-lying areas that slow down floodwaters and can hold a significant volume of water, mitigating damage downstream (Watson et al., 2016). 4. Overbank deposition is the process whereby sediment in flood waters settles out of the water column and onto the tops of river banks, floodplains, and other surfaces that are inundated. 5. Point bar deposition and in-channel deposition are processes of sediment settling out of the water column onto point bars (at the inside edge of a curve, bend, or meander, in the river) and channel bottoms, often occurring during, but not limited to high-water events. 6. Bank erosion occurs both by water-driven sediment removal from channel edges and by gravitydriven slumps that can extend above the water elevation, especially during high water events that undercut banks. Bank erosion typically occurs on the outside edge of a river bend. 7. Avulsion is the process where an old channel is rapidly abandoned and a new channel is created, often cutting off a river bend. In addition to cutoff avulsions, alluvial fan avulsions can occur when rivers create new, steeper paths down alluvial fans, which are cone-shaped deposits of sediment that can build up where river channels (or valley floors) have a marked decrease in slope.

10 8. In-channel erosion occurs when moving water mobilizes sediment, mostly during high water events when transport thresholds are exceeded and the water has enough energy to carry the sediment away. 9. Near-channel landslides are similar to bank erosion and slumps but larger and can extend several meters above the elevation of flood waters on steep land adjacent to channels and floodplains. Steep terrain processes such as surface runoff, creep, mass wasting and litter fall can occur adjacent to river channels. These processes create accelerated delivery of sediment, wood, and other materials to rivers even in the absence of landslides. Finally, riparian ecologic processes occur within and adjacent to the channel, floodplains, and nearby wetlands, forming a unique transitional ecotone between wetland and upland areas. Riparian vegetation adjacent to channels delivers carbon, provides shade, protects banks from erosion, traps sediment, consumes and stores nutrients and pollutants, and creates complex habitat at channel margins (Hupp and Osterkamp, 1996; Naiman and Decamps, 1997; Wohl, 2017). Similar to the wetland processes, we feel that the ecological services, flood-wave mitigation, and bank-stabilization that these riparian zones provide to river systems are of critical importance, and should be explicitly included in our corridor. 2.2 Step 2 Definition of Process Units Because many of these river processes occur in overlapping areas, are driven by similar mechanisms, and operate in similar ways within the fluvial environment, many of these processes can be grouped together. In this step, we organize the processes into five units that can be mapped together: a) flood processes, b) landslide and steep terrain processes, c) wetland processes, d) channel migration processes, e) riparian ecologic processes (Table 1). Photos depicting examples of these process groups are shown in Figure 2. a) Flood Processes Unit. This unit intends to capture the many processes that occur only where flood waters reach. These processes include flooding, overbank deposition, point bar deposition, in-channel erosion, and in-channel deposition. The Flood Processes Unit also generally overlaps with the area of modern alluvium, which is loose granular sediment deposited by rivers in channels, point bars, and floodplains. Alluvium is unconsolidated, and therefore significantly more erodible than bedrock. Therefore, the Flood Processes Unit also characterizes most, but not all, of the areas likely to experience bank erosion, channel migration, and channel cutoff avulsions. (Figure 2a. These correspond exclusively to 1, 4, 5 and 8 from Figure 1, and also overlap with 2, 3, 6, and 7). b) Wetland Processes Unit. This unit intends to capture near-channel wetland processes that may exist within the extent of the Flood Processes Unit, but may also extend above and beyond floodwaters. Specifically including these critical wetland habitat areas within the river corridor is important both for floodwater dissipation, as well as the intrinsic value and ecosystem services they provide. (Figure 2c. These correspond exclusively to 3 from Figure 1). c) Steep Terrain and Landslide Processes Unit. This unit intends to capture steep terrain and nearchannel landslides that can extend beyond floodwaters and up onto hillslopes and steep areas near streams. These steep terrain processes include overland runoff, creep, mass wasting and litter fall on

11 steeply sloping terrain adjacent to channels and flood-prone areas. (Figure 2b. These correspond exclusively to 9 from Figure 1, and also overlap with 3 and 6). d) Migration Processes Unit. This unit intends to capture channel migration, bank failures, and cutoff avulsions, which occur predominantly in the floodplain but may also extend laterally into terraces and other granular deposits. Most channel migration occurs within the bounds of the flooded area due to the low topography and erodible alluvium. However, erodible material, often Quaternary glacio-fluvial deposits, may exist adjacent to the floodplain. Thus the Flood Processes Unit is not sufficient to include all areas prone to these migration processes, and a Migration Processes Unit is warranted. This unit may miss some areas prone to avulsions on alluvial fans. We recognize the importance of alluvial fan avulsions, but they are rare on rivers in the North Atlantic region. They occur more commonly in large mountain systems along low order streams in the western US. The refinements described in Step 5 allow for an alluvial fan process unit at specific locations if desired. (Figure 2d. These correspond exclusively to 7 from Figure 1, and also overlap with 2 and 6). e) Riparian Ecologic Processes Unit. This area intends to capture the ecologic processes in the riparian zone. The area of these ecologic processes is not entirely captured by the process units described above, especially along rivers with flood control and in steep, narrow valleys. In locations where flood control has limited the extent of modern floodwaters, vestigial riparian vegetation communities may exist and provide critical riparian habitat. In steep and narrow valleys, riparian vegetation may extend above and beyond the flood width but still deliver organic matter to channels efficiently and create habitat. Much of the riparian vegetation in river corridors exists within the four other process units, but because riparian vegetation plays such an important role in river corridor and can extend beyond the edge of the other process units, we include it as a separate process unit. (Figure 2e. These correspond exclusively to 3 and 9 from Figure 1, and also overlap with 3 and 6). Table 1. River Processes included in the process unit as part of Step 2, PURE River Corridor delineation Process Units 2 Process 1 Flood Steep/ Wetland Channel Riparian Landslide Migration Buffer 1 Flooding O 2 River migration P O 3 Wetland processes P O P 3 Riparian ecologic processes P P O 4 Overbank deposition O 5 Point bar deposition O 6 Bank erosion P P O P 7 Avulsion P P 8 In-channel deposition O 9 Near-channel landslides O 9 Steep terrain processes O O Notes: 1 numbers correspond to river processes shown in Figure 1. 2 Processes that occur only within a given process unit are marked with the letter "O" Processes that occur partially within a process unit but can also occur beyond the region of a process unit are marked with the letter "P".

12 2.3 Step 3 Mapping rules In this step, we detail the rules for mapping the five process units. This procedure is modular, such that each of the designated process units is independently mapped and summed together with the others. The procedure is also iterative: if an improved data set or mapping procedure is developed for a process unit or an individual process, then it can be swapped for the existing rule, remapped, and the resulting area can be summed with the other process unit areas. The challenge that we attempt to address in this step is to choose rules that faithfully represent the five process units and accurately map all of the processes we identified in Step 1 with data that is widely available, such that river corridors can be delineated across a large geographic region. This may mean that we are not using the highest-resolution data available (for example, a detailed, reach-scale investigation of the extent of a riparian species of concern, or high-resolution LiDAR elevation data), but it does guarantee that data are consistently available and similar across the region. If refinement is desired for a limited-geographic-scope study, these data could be substituted in, as described in Step 5 Refinements. In general, we favor simple over complex rules, and physically-based rules over empirical rules wherever possible. The protocols also avoid hand digitizing, to allow reproducibility and mapping over large areas. 2.3 a) Flood Processes Unit mapping rule: The Flood Processes Unit mapping rule is to identify areas subject to inundation. We sought a data set with a physically based representation of areas that are likely to flood over century time scales, which aligns with human time scales on the order of 1 to 100 years. For our purposes, the Federal Emergency Management Agency (FEMA) provides the most widely available geospatial data of flood-inundation areas from hydrologic and hydraulic modeling. The 1% Annual Exceedance Probability (AEP) flood extent shows the elevation and width of a flood that has the 1% probability of occurring in any given year at a given location. We ve chosen the 1% AEP flood for the Flood Processes Unit because it fits with human time scales, and also takes advantage of a common vernacular used for planning purposes among river scientists, engineers, and land managers. The 1% AEP flood is sometimes misunderstood by non-specialists. It is often called the 100-year recurrence interval flood, but having a 1% AEP flood in one year does not indicate that it will be 100 years until the next 1% AEP flood at that location. Considering broader areas instead of a single location, there is a high likelihood that a 100-year flood will occur somewhere in the Northeast region every year. There is always uncertainty in delineating areas prone to inundation. Uncertainty arises from sporadic river gaging records, assumptions in hydrologic and hydraulic models, and the resolution of topographic data (Bales and Wagner, 2009). In addition, the magnitude of the 1% AEP flood may increase with landuse change, such as increased impervious area from roofs and roads. Landuse changes, however, have less of an effect on large magnitude floods like the 1% AEP flood than on moderate floods, like 1.1- to 15-year recurrence interval flood (Pitlick, 1997; Rose and Peters, 2001). Climate change may also adjust the magnitude of rare flood events, and this is an active area of research that, in short, shows that hydrologic conditions are changing over time and risk of floods is increasing (Milly et al., 2008; Milly et al., 2002; Rawlins et al., 2012; Yellen et al., 2016). Nonetheless, the 1% AEP flood provides critical information for predicting processes that will occur in the future and have rarely been observed in the past. We recognize that the 1% AEP flood area does not

13 equate to the entire river corridor. Thus, the PURE method includes four other process units to capture the full suite of river corridor processes. Procedure: We use the FEMA Special Flood Hazard Area (SFHA) in the National Flood Hazard Layer (NFHL) database ( and extract the polygons labeled as T in SFHA_TF column of the attribute table of the S_Fld_Haz_Ar layer. These polygons equal the Flood Processes Unit area. The widespread availability of FEMA maps is a major advantage for consistent mapping of river corridors across large regions. Figure 3 shows the availability of FEMA flood data across the NALCC region. One limitation is that FEMA maps were not created for all areas. For example, roughly the western half of Massachusetts is not covered. Even in regions that are covered, the FEMA flood data typically do not show flood areas for small rivers, such as 1 st and 2 nd order streams. A second limitation is that the accuracy may be low when conditions have changed since maps were created. FEMA maps are being continually refined and improved by FEMA and private consultants. Future Work, Model Refinements, or Other Data Sources: In addition to FEMA flood data, other options exist to determine probable flood extents. The USGS is creating flood-inundation maps in cooperation with FEMA in selected reaches (Bent et al., 2015; Lombard and Bent, 2015) as part of the Flood Inundation Mapping Program. These modeling efforts utilize high water marks from recent large floods, field-surveyed cross sections, and high-resolution topography from LiDAR to make very accurate maps of inundated areas in recently-flooded locations. Another modeling option is to use HEC-GeoRAS, a hydraulic modeling program developed by the U.S. Army Corp of Engineers, in places where LiDAR data are available using a method similar to the recent USGS efforts but without field data of the underwater stream bed or high water marks. The modeled results are less accurate than the recent USGS efforts, with likely wider modeled flood areas. The procedure is not applicable for regional scale analysis because it is too time intensive to achieve suitable results, nor is it possible in areas without LiDAR data (Gartner et al., 2016). The accuracy is also not acceptable for the standards of FEMA maps that establish flood insurance rates. Yet the accuracy can be suitable for the planning purposes of reach-scale river corridor delineation at sites of interest. Another emerging option is to extract floodplain extents from topographic analysis of LiDAR data. Floodplains can be identified using automated extraction programs in areas with relatively flat surfaces that are slightly higher than channel elevations (Clubb et al., 2017; Samela et al., 2018). Initial results show good agreement with FEMA mapped flood areas. The limitation of this option is that LiDAR data are likely forthcoming (Massachusetts recently released data for the entire state) but not yet available to the public for the entire NALCC region. 2.3 b) Steep terrain and landslide processes mapping rules: This rule aims to map all river-adjacent land areas (within reasonable bounds) with slopes steep enough for rapid transport of materials downslope, using a slope angle that is a conservatively low estimate of the angle of repose, angle of internal friction, and the typical failure surface angle of landslides. Procedure: Include topography with a slope greater than 20 degrees and within 60 m of the Flood Processes Unit. The data source for topography is the 1/3 arc-second Digital Elevation Models (DEMs), approximately 10-m grid size, from the USGS National Map (Gesch et al., 2002).

14 We set the 60 m distance based on observations of near-channel landslides triggered by Tropical Storm Irene in Vermont and Massachusetts (Dethier et al., 2016; Gartner et al., 2015). We did not aim to capture every landslide across the entire landscape, but rather, specifically those landslides that were thought to have been caused by bank-undercutting, slope toe erosion, and other similar river-driven processes. Some steep slopes adjacent to channels extend continuously beyond 60 m, but the channelproximal locations have more rapid and efficient delivery of materials than distal locations. The procedure fits with the guiding principles of being physically-based, simple, and easily reproducible. Future Work, Model Refinements, or Other Data Sources: One shortcoming of this rule as currently configured is that steep areas are only captured if they are twice as wide as the resolution of the DEM, which would be more than ~20 m wide when using ubiquitous 1/3 arc-second DEMs from the USGS. As a result, the mapping rule may lead to missing some areas susceptible to landslides and other steep terrain processes. If the corridor were to be mapped for a smaller geographic area, however, where LiDAR or other high-resolution terrain data were available, a more granular representation of these steep terrain processes could be captured. On the other hand, because there is some overlap in the physical spaces where different processes occur, the Riparian Ecologic and Channel Migration Processes Units are likely to also capture the steep terrain processes in areas less than 20 m from the channel. Other more complex rules are available for mapping steep terrain processes. For example, stream power gradient analysis of large storm events shows that landslides and other steep terrain processes are more likely in river reaches that have downstream increases in stream power, modeled in GIS from 10-m digital elevation models (Gartner et al., 2015). An additional refinement could be to add a layer of analysis to the steep processes mapping rule to only include areas with downstream increasing stream power. Another example of a more complex algorithm is a GIS-based model that predicts landslide locations in Massachusetts based on slope angle, soil properties, the angle of internal friction, cohesion, and wetness (Mabee and Duncan, 2013). Some elements of this work could also be included in this mapping rule. Even though the GIS-land failure model is potentially more accurate than the simple approach that we use here based on slope and distance alone, it was not parametrized for regions outside Massachusetts (originally developed for North Carolina), and lumps many values into a single, weighted parameter, reducing the method s flexibility. It may also give a false sense of accuracy for other regions within the NALCC, especially locations with different geology and glacial history. 2.3 c) Wetland processes mapping rule: This rule is intended to include all known and mapped wetlands adjacent to the river within a reasonable distance. Procedure: Include all wetland polygons in the National Wetland Inventory (NWI) database (Wilen and Bates, 1995) that intersect within 5 m of the flood process unit or intersect within a 5 m of the channel. The channel locations are obtained from the National Hydrography Dataset Plus High Resolution (NHDPlusHR) dataset (McKay et al., 2012). Specifically, we use the Wetland layer from the wetlands database, and the NHDflowline and NHDArea layers from the NHDPlusHR geodatabase. The NHDPlusHR dataset provides geospatial data of channel centerlines and channel areas, available for download at ( The original source is the USGS 1:24,000-scale printed topographic maps, which have a locational accuracy standard requiring ninety percent of well-defined features to lie within 40 ft of their true geographic position based on the date of collection. As with any

15 large dataset, these data are not a perfect representation of the drainage network, but ongoing efforts to publish NHDPlusHR data across the U.S. have improved the accuracy of these products. NHDPlusHR data mimic the depiction of streams and rivers of the printed topographic maps. Blue lines show centerlines of all streams and rivers. Blue polygons show the channel extent only of rivers greater than a few meters wide (such as the Connecticut River, as shown in Figure 5c). The buffer is applied starting from the channel edge for larger channels where the NHDPlusHR dataset depicts the channel area, and from the channel centerline for channels less than a few meters wide where the NHD dataset depicts that channel location only as a line. Because the buffer we use is applied to the centerline for very small streams, not the channel edge, this could potentially underestimate the buffer by a few meters. The NWI is a digital database of wetlands developed and maintained by the USFWS, available for download at Wetlands were initially mapped for most of the coterminous U.S. using mid-1980 s color infrared photos from high-altitude aerial photography (Cowardin et al., 1979). Digital data of wetland locations are available for 81% of the U.S, and 100% of the continental U.S. The NWI is now being updated at a rate of 2% per year. Future Work, Model Refinements, or Other Data Sources: As with any digital dataset of land surface characteristics that cannot be directly observed, there are errors and inaccuracies. Local wetland delineations are often conducted for land use permitting and may include more accurate data or detail than the NWI database. However, for broad-scale corridor delineation, the NWI is a fortuitous and broad-reaching dataset. For the reach scale corridor delineation, it may be preferable to use more detailed wetland map data based on field observations of soils, hydroperiods, and vegetation. 2.3 d) Channel migration mapping rules: The mapping rule for channel migration is to capture an area large enough to accommodate present and future channel migration. The challenge for the channel migration process unit is that it is extremely sensitive to the condition (e.g. channel location) that comes immediately before the prediction, and becomes less accurate with time after that. For the present version of this mapping rule, we propose that most channel migration occurs predominantly in the alluvial area, which also corresponds to the Flood Processes Unit. But channels can be close to the boundary of the flood unit, perhaps abutting the edge of historical river terraces that are high enough not to be mapped as flood-prone, but still comprised of easily erodible and/or unconsolidated material. A margin of safety is therefore needed to help capture this type of bank erosion (particularly prevalent in incised streams) and migration processes beyond the flood zone. To do this, we take the mapped channel edge, and add an additional buffer to it. The buffer plus the Flood Processes Unit captures the likely meander migration zone. Procedure: Compute a 40 m buffer from the channel edge in the NHD dataset, and combine this area with the area of the Flood Processes Unit to delineate the Channel Migration Processes Unit. The channel edge in the NHD database is described in the Wetlands Process Unit rule. Future Work, Model Refinements, or Other Data Sources: This rule is guided in part by the goal of keeping mapping rules simple and as straightforward as possible, so that corridors can be mapped efficiently over large regions. River migration is a very complex process to predict, with a large range of migration rates in different settings. Rates of average bank retreat range from 0.0 to 7.3 meters per year in locations in the U.S. and Europe (as observed by Knighton, 2014). The rate of river migration is influenced by channel curvature, channel width, erodibility of adjacent material (consider bedrock versus sand, as well as variability in riparian vegetation), and the sequence of storm events (Buraas et

16 al., 2014; Hickin and Nanson, 1984; Schwenk et al., 2017). Not only are these factors highly variable naturally in any given watershed, but also these factors are highly influenced by human activities that straighten channels, reinforce banks, remove riparian vegetation and woody debris, and curtail flood flows. If we then add within-watershed variability, and climate change to the equation, the complexity of river migration quickly gets computationally expensive and infeasible. While in theory it is possible to use existing bank retreat rate data to determine and model likely or potential meander migration, the process of doing so may be inordinately time consuming, data intensive, and poorly constrained. One might be tempted to use a maximum known migration rate as a worst case scenario, but it would not be appropriate to compute a channel migration zone by assuming that rivers could migrate laterally at a rate up to 7.3 m per year. This approach would equate to 730 m swath on each side of the channel over a century time-scale! Rivers do not simply move laterally in the same direction indefinitely. Instead they meander back and forth, typically across the zone of modern alluvium. In addition, as noted above, the best predictor of where the channel will move next is where it is now and over 100 years, that position (and our ability to predict it) can change significantly (see Figure 6). Other approaches to predict areas of channel migration exist, but they are better suited for analysis of a single reach or a single river rather than region-wide predictions. Recent studies have used channel curvature together with channel width and erodibility of adjacent materials to predict the rate of river migration (Buraas et al., 2014; Schwenk et al., 2017). This approach can produce accurate predictions, but it is data intensive and highly dependent on initial conditions. Another approach is to assume that future channel migration will follow in a similar area as past river migration, and use historic maps of former channel locations to estimate a historical meander zone. As part of the development of the threshold values for this rule, we explored using a different algorithm for determining the buffer width. Initially, we applied the 40 m buffer because it mirrored the Massachusetts Clean Water Act buffer width for pristine streams and effectively captures large regions surrounding headwaters, which are important upland habitats. That said, we recognize that 40 m might be disproportionately large for very small streams, and so we experimented with a buffer that scaled up with increasing channel width. Unfortunately, channel width estimation also presents challenges an potential inaccuracies. Channel width can be estimated from regional curves that relate drainage area to channel width. Deciding which regional relationship to use, and how to flag an automated program to use the correct equation presented a problem, since there are more than 20 published studies on regional relationships across the 13 states in NALCC area (Bieger et al., 2015). As we discovered, the analysis quickly became more complex and computationally intensive without clear improvements in accuracy, so it was abandoned, but could be revisited at a later date. Other approaches to determine the likely zone of channel migration exist that utilize channel curvature, channel width, erodibility of adjacent material and/or critical stress calculations to predict channel movement. But, neither of these are easily automated, independent from initial conditions, or widely applicable across a geographically diverse region. 2.3 e) Riparian ecology mapping rules: The mapping rule for the Riparian Ecologic Processes Unit is to capture the unique riparian zone ecology by adding a buffer to the existing mapped river channel edge. This step allows practitioners a particular process unit within which to define a parameter of interest and include it in the total PURE corridor to suit their specific needs. Here, we define the riparian zone as

17 the vegetated strip adjacent to channel and floodplains comprised of diverse and complex habitat for many species. This zone also helps stabilize river banks, slow down high flood flows, and filter pollutants through overland flow. Procedure: Add a 40 m buffer from the channel edge (the channel edge as defined by the NHD database is described in the Wetland Processes Unit rule). This buffer is combined with the Flood Processes and Wetlands Processes Units to create the Riparian Ecologic Processes Unit. Natural and human-altered waterways exhibit great variability in the distance that riparian vegetation extends from the edge of channels and floodplains, and the value of 40 m is supported by studies that have examined or proposed riparian buffers of 10 to 40 m (Castelle et al., 1994; Fischer and Fischenich, 2000; Lee et al., 2004; Micheli et al., 2004). Other existing river protection programs use a variety of setbacks or buffers, ranging from 7 to 22 m, including the USDA Conservation Reserve Enhancement Program (10 to 55 m), Massachusetts River Protection Act (15.2 to 61 m), Vermont Small Stream Setback (15.2 m) the Wild and Scenic River Act (75 to 122 m). We recognize that riparian ecologic processes are diverse, and the locations of these processes are difficult to predict and/or data and time intensive to measure. In practice, this mapping rule ensures a minimum 40 m buffer on all streams, even if there are no wetlands, steep areas, or flood extents mapped along a given reach. This may not capture the extent of ecologic processes perfectly, and there is the possibility that riparian ecological areas exist outside of the 40 m buffer from the channel, but we must select a value for practical purposes, so we ve chosen a very inclusive value, one of the larger included in published studies. Future Work, Model Refinements, or Other Data Sources: More sophisticated, and potentially more accurate, methods exist for designating the location of crucial, existing riparian species. Detailed field studies of specific species, remote sensing analysis, and ecologic modeling can be conducted at specific sites of interest. For example, the USFWS has developed methods for delineated riparian areas in arid regions based on vegetation abundance (USFWS, 1997) but these strategies are not as applicable in humid areas with more abundant vegetation in both riparian and upland areas. 2.4 Step 4 Overlay process units Overlaying the process units is a straightforward step completed in ArcGIS or other GIS programs. The work flow is depicted in Figure 4. Essentially, the channel area, flood area (a), landslide area (b), wetland area (c), migration area (d), and riparian ecologic area (e) are combined into a single polygon that is the sum, or union, of all five of these areas. We smooth the result in ArcMap by applying a 40 m buffer to the total river corridor and then applying a -40 m buffer. This smoothing step effectively smooths out all of the outward-facing pixelated crenulations along the boundaries. The resulting corridor boundary does not appear to be excessively detailed, so as not to imply a greater level of precision than the corridor possesses warrant. In addition to generation of the total corridor, we preserve each of the process unit s polygon layers. This way, a user can investigate layers individually, weight the layers differently, or use some as regulatory layers and others as advisory layers for river corridor management. 2.5 Step 5 Refine as necessary for desired use Several options exist to refine the PURE delineation method, and these fall roughly into two categories: (i) altering individual rules and (ii) adding process units.

18 Altering individual rules: Within each of the Process Units is a discussion of Future Work, Model Refinements, or Other Data Sources which addresses this mode of refinement. Some of these alterations include using a different, or more precise data source. Others include using an improved model, different equation, or updated methodology for determining that unit. And finally, the least substantial alterations include changing the distances or thresholds used in the GIS algorithm to make the maps. Rules may be altered for site specific river corridor delineation. We ve tested this method across a wide range of example watersheds and selected the parameters and thresholds that work best across the northeast region. However, given some site specific information to test against and a limited geographical range, these could be adjusted to reflect the most accurate representation of the conditions at that precise location. For example, detailed river migration studies along a reach could be substituted for the existing river migration mapping rule. This substitution may improve the accuracy of the river corridor, but this map would now be site-specific and not uniform if re-integrated into regional maps. Or, in areas where FEMA flood hazard mapping layers do not exist, hydrogeomorphological units based on surficial geologic maps could be used in place of these. Adding process units: The PURE method is not limited to the five process units included here. If land managers were to have a heightened interest in, for example, depicting the habitat for riparian birds within river corridors, a riparian bird movement unit could be added. Others may have heightened interest in pollutant transport or geochemical cycling in river corridors. In such cases, a pollutant transport processes unit or a geochemical cycling processes unit could be added with accompanying mapping rules. The delineation procedures we present here do consider these processes, but not in great detail. In addition, while not explicitly river processes, there may be a desire to overlay different historical, ecological, recreational, or cultural resources maps with the river corridor. These could be derived according to their own sets of mapping rules, and then added to the total corridor for land management and resource protection. 3. Test Studies The process unit river corridor delineation was performed in three locations across the NALCC to evaluate the delineation method s effectiveness at capturing the desired river processes, to exhibit how individual process units compare with field evidence from flood events, and to illustrate the differences and similarities between the PURE Corridor and other widely-used corridor delineation methods. 3.1 Example 1 Northern Connecticut River Valley: This example is used to investigate the layering of the five process units that comprise the PURE Corridor, to compare the Migration Processes Unit with evidence of past channel movement, and to compare the PURE Corridor with other corridor delineation strategies. The site is a mixture of forest and farmlands, with rolling topography on both sides of the broad Connecticut River valley situated in Haverhill, NH and Newbury, VT (Figure 5). A mixture of 1 st to 4 th order tributaries lead to the 7 th order Connecticut River, which forms the boundary between Haverhill, NH and Newbury, VT. The area was glaciated during the last glacial maximum, and the valley bottom was occupied by post-glacial Lake Hitchcock from approximately 15,600 to 12,900 B.P. (Ridge and

19 Larsen, 1990). The methods to map and layer the five process units follow the standard procedures without refinements, as described in Section Process units that comprise the PURE Corridor The Flood Processes Unit (Figure 5d) is equal to the FEMA 1% AEP flood extent. It varies in width along the Connecticut River and is discontinuous in smaller tributaries, with no flood extent mapped for the smallest streams. The Flood Processes Unit is discontinuous in part because it was derived from the FEMA 1% AEP flood extent map data, which are discontinuous, and in part because the smallest streams have little space for broad inundation and pooling near their origins in steeper areas. The Landslide and Steep Processes Unit (Figure 5d) is derived from slope analysis of 1/3 Arc-second DEMs and includes hillslopes greater than 20 degrees that occur within 60 m of the Flood Processes Unit. This encompasses terraces on the eastern side of the Connecticut River and other steep hillslopes scattered throughout the view. The unit does not include steep hillslopes adjacent to the smallest tributaries, due because of discontinuities in the flood process unit. However, locations of potential steep processes adjacent to small tributaries are effectively included in the total corridor through other overlapping process units such as the channel migration and riparian areas. The Wetland Processes Unit (Figure 5e) was derived from the NWI and includes mapped wetland areas within 5 m of the flood process unit or within 5 m of the NHD mapped drainage network. The NWI data generally show present-day channels, for example the present-day Connecticut River channel seen in this figure. The Migration Processes Unit also shows narrow sinuous wetlands in some swales that do not have mapped channels in the NHD dataset. Some of these appear as narrow spurs in the total river corridor (Figure 5h). Many wetlands within the NWI are in the uplands and do not intersect within 5 m of the channel or Flood Processes Unit. While important for ecological processes of wetlands in the overall watershed, these upland wetlands are not included in the corridor because they do not regularly and directly interact with the river nor influence river processes on the human timescales that are the focus of this river corridor delineation effort. The Migration Processes Unit (Figure 5f) is the sum of the Flood Processes Unit plus a 40 m buffer from the channels in the NHD dataset. The buffer is applied starting from the channel edge for larger channels where the NHD dataset depicts the channel area, and from the channel centerline for channels less than a few meters wide where the NHD dataset depicts that channel location only as a line. The white arrow in this panel shows a location where the Connecticut River channel is adjacent to the eastern margin of the Flood Processes Unit, and the 40 m buffer expands the width of the total corridor in this location. The 40 m buffer of the Migration Processes Unit also expands the width of the total corridor along the smaller streams which do not have flood-prone areas mapped by FEMA. The Riparian Ecological Processes Unit (Figure 5g) is the sum of the Flood Processes Unit, the Wetland Processes Unit, and a 40 m buffer on the channel edge. In some areas, such as along the Connecticut River, the 40 m buffer is less than the flood-prone area, and it does not add to the width of the corridor. In other areas, for example small tributaries without wetlands or FEMA flood-prone areas, the 40 m buffer on the channel edge is as wide as or wider than any other unit. The total corridor is the combination of all 5 process units, which we also refer to as the PURE Corridor (Figure 5h). At a minimum, the PURE corridor is 40 m wide in all locations with channels mapped in the NHD due to the 40 m minimum of the Migration Processes and Riparian Ecological Processes Units.

20 3.1.2 PURE Corridor compared to past channel migration. There seems to be agreement between the channel Migration Processes Unit and the former locations of the Connecticut River. Historic maps and modern aerial imagery show from 0 to 90 m of channel migration in the reach between 1935 and the present. However, the NWI data (Figure 5e) show oxbowshaped wetlands along the Connecticut River Valley that were likely former channel locations, suggesting between 250 and 1000 m of lateral movement. Scroll bars are evident in the aerial imagery (Figure 5b) between some oxbow-shaped wetlands and the modern channel. The spatial domain of the Migration Processes Unit (Figure 5f) encompasses 100% of these features in this 15 km segment of the Connecticut River Valley. The overlap is evidence that channel migration has occurred primarily within the floodplain, within the erodible alluvium. The Migration Processes Unit is satisfactory in this example, to the extent that past conditions are the key to present and future conditions. Regarding future conditions, this site also exhibits areas where the 40 m buffer in the Migration Processes Unit is intended to account for possible migration into erodible terraces which may lie above the elevation of the flood-prone areas mapped FEMA (white arrow, Figure 5f) PURE Corridor compared to other River Corridor delineation strategies. This northern Connecticut River Valley site is also used to compare the PURE Corridor with other delineation strategies. For the Vermont Corridor, we merged the Vermont state River Corridor shapefile with the Small Stream 50-ft Setback shapefile, both downloaded from the state of Vermont ( Note that the Vermont Corridor is only completed to the west of the Connecticut River. The state has not yet developed a strategy to delineate the corridor along the Connecticut River, and areas to the east are in New Hampshire. In this view (Figure 5i), we see that the Vermont Corridor covers fewer streams than the PURE Corridor. There are also gaps along some small streams, in part because the Vermont Corridor does not delineate streams with drainage areas less than 0.25 mi 2 and in part because the Vermont Small Stream 50 ft Setback shapefile was released in July 2018 and is still being refined. The Nature Conservancy s ARA corridor is broadly consistent with the PURE Corridor, but there are noteworthy differences in both the reach-scale analysis and the physical rationale behind inclusion of certain areas in the corridor (Figure 5j). The ARA spatial data was downloaded from While the ARA method could likely be applied on any resolution of spatial data, a coarse 30-m DEM was used to create these publicly available data. In this view, the ARA is more pixelated and shows isolated small groups of pixels. With this pixilation, the ARA is better suited for regional, rather than reach-scale analysis. It is not clear why some of the isolated patches are included in the ARA beyond knowing that these patches must be the result of the cost surface, or distance-slope equation, being higher than the pre-determined threshold value. Compared to the PURE Corridor, the ARA tends to cover a slightly narrower width along Connecticut River main stem, broader width along smaller tributaries, and less total length along small tributaries. Several areas are not included in the ARA but are included in the PURE Corridor along smaller tributaries and adjacent to the Connecticut River main stem. We cannot conclude which method is over-predicting or under-predicting the corridor in different areas without direct observations of river corridor processes; however, we can at least provide a physicallybased explanation and a transparent rationale for why areas are included in PURE Corridor. For example, a white arrow (left side of Figure 5j) near the Connecticut River shows a broad area in the

21 PURE Corridor because it is within the area of the Flood Processes Unit depicted in Figure 5d and is susceptible to flooding via inundation. This area is not in the ARA simply because this area does not exceed the pre-set threshold value. As another example, a small area is included in the ARA as a spur from a wider corridor along one of the eastern tributaries (white arrow, right side of Figure 5j). It is unclear why this might be an area of interest in the ARA result. But by examining the component layers, or Process Units, from the PURE corridor, we can easily see that this a wetland (Figure 5g), and therefore of interest for the wetland-specific processes present in this location. Figure 5k compares the equivalent of the Massachusetts River Protection Area with the PURE Corridor. A 200 foot buffer was applied to the edge to the NHDflowline and NHDArea layers from the NHD geodatabase to simulate Massachusetts program in this area of Vermont and New Hampshire. Both the River Protection Area and the PURE Corridor cover every known stream in this view. But the PURE Corridor is wider in some areas, for example, along the Connecticut River valley in areas that have flood hazards, susceptibly to river migration, wetland processes, and important riparian habitat (white arrow, Figure 5k). The River Protection Area underestimates the width of the river corridor in these locations. 3.2 Example 2 Baker River: This site (Figure 6) is chosen to assess the accuracy of the Migration Processes Unit in an active setting where the channel position changes frequently. The site is located along a ~1.5 km reach in Warren, NH where the Baker River transitions from the steep headwaters on the flanks of Mt Moosilauke to less steep terrain in a broader valley. In some places the river is braided, an indication of high sediment loads. The site has experienced several high water events in recent years, notably during Tropical Storm Irene in 2011, and during a storm on October 30, In the 2017 storm, channel migration undercut the eastern bank below a home, and carried it downriver to where it was destroyed when it smashed into a bridge. (Video available online at For methods, channel locations over the last 25 years were digitized from historic imagery available from Google Earth (1993, 2003, 2011, and 2013) and ESRI (2015). To delineate the PURE Corridor, we used the standard procedure without refinements, as described in Section 3. These methods have no refinements based on site-specific information. The data show that despite rapidly changing channel positions, all known former channel locations are located within the spatial domain of the Migration Processes Unit of the PURE Corridor. Through most of this study site, the channel positions are located within the predicted flood area from FEMA, evidence that channel migration occurs primarily within the floodplain. However, highly erodible terraces and alluvium may fall outside of the FEMA area. For this reason, it is noteworthy that the PURE Corridor captures some of these vulnerable reaches, including a 100 m segment (white arrow, Figure 6) where recent, former channel locations are beyond the limit of the FEMA flood area. If a site-specific river corridor delineation were undertaken along this reach, we might refine the mapping rule for this unit, as discussed in Step 5 of the PURE methodology. A possible refinement would be to apply the 40 m buffer to the most recent known channel position (rather than whichever channel position is in the NHD database) as ascertained from drone photogrammetry, satellite imagery, or field

22 surveying. Because the channel position changes so frequently at this location, it may warrant special notes for frequent updates resulting from recent storms that may intensify under climate change. 3.3 Example 3 White River Watershed: In this location we examine how the PURE Corridor compares with evidence of geomorphic change during a major storm, Tropical Storm Irene in Since the PURE corridor is created using data available before Irene, we can evaluate how well our method predicts areas that are susceptible to natural hazards and other important processes of material transport during the flood. The site is also used for quantitative comparison of different strategies to delineate areas of concern near rivers, including the Vermont Corridor, ARA, FEMA flood area, and Massachusetts Rivers Protection Act area. We do not expect perfect predictions of the hazards from any strategy, especially for a rare and extreme event. Yet the different strategies can be evaluated based on how well they encompass areas susceptible geomorphic change, balanced by the total area they cover. The White River watershed, has the largest free-flowing river network in Vermont, with a combination of forest and farms, wide and narrow valleys, and small towns among the Green Mountains (Figure 7). In August 2011, Tropical Storm Irene produced abundant floodplain sedimentation, hundreds of landslides, and record breaking high flows at many gaging stations in Vermont (Magilligan et al., 2015). Floodplain deposits and landslides from Irene were mapped by Gartner et al. (2015) along ~ 50 km of the White River and the West Branch of the White River, and landslides from Irene were mapped by Dethier et al. (2016) across much of Vermont and Western Massachusetts. The study area examined here is the southern 955 km 2 of the White River watershed for which FEMA flood data are available. For methods in this example, the PURE Corridor was delineated using the standard procedures without refinements, as detailed in Section 3. As in the previous test sites, no site-specific information was used to delineate the PURE Corridor. The ARA extent was downloaded from the Nature Conservancy ( and the Vermont River corridor data was downloaded from the state of Vermont ( As in the previous example, we merged the Vermont River Corridor shapefile with the Vermont Small Stream 50-ft Setback shapefile. We replicated the Massachusetts River Protection Act area by applying a 200 foot buffer to the edge to the NHDflowline and NHDArea layers from the NHD geodatabase. Across the entire study area, the PURE Corridor overlaps with 86% of the area of floodplain deposits and 92% of the area of landslides (Table 2). These data show that the PURE Corridor is a successful predictor, even in an extreme event, of a majority of the areas prone to river-related geomorphic processes. These geomorphic processes mobilize materials, shape the landscape, and sometimes create natural hazards that threaten roads, agricultural land, and structures. Figure 7 shows a detailed view of a representative portion of the study area, where the West Branch of the White River meets the main stem of the White River in Rochester, VT. The sediment deposited on the floodplain during Irene is shown in yellow, and landslide areas are shown in red. The PURE corridor encompasses most, but not all, of the areas with floodplain deposits and landslides (Figure 7c). By comparison, the other strategies have about equal success in predicting areas of floodplain deposition, but varying degrees of success in identifying locations prone to landslides. The Vermont Corridor, ARA, FEMA floodplain, and Massachusetts Rivers Protection Act areas overlap with between 82% and 91% of the floodplain deposit areas (Table 2, Figures 7d, 7e, 7f, and 7g). The MA River Protection Act is a notable exception; it overlaps with only 53% of floodplain deposits. Regarding landslides, the ARA and MA River Protection Act show high success in predicting areas susceptible to

23 landslides (99% and 96% overlap, respectively), but the Vermont Corridor and FEMA flood area overlap with only 60% and 15% of the landslide area, respectively. This is not a direct criticism of either the Vermont Corridor or FEMA, as these delineations were not intended to cover areas prone to mass wasting processes, such as landslides, into drainage networks. Nevertheless, the results show that FEMA flood areas alone are an inadequate tool for predicting the geomorphic hazards associated with flooding. Some of the successes of these strategies arise from the different total areas that they cover. For a hypothetical end-member example, a corridor that is thousands of meters wide could cover 100% of areas susceptible to river-related geomorphic processes, but this would not be helpful for directing limited funds for river protection and management. Ideally, a river corridor covers the minimum area necessary to encompass all of the locations susceptible to geomorphic hazards within the river corridor. The ARA covers the largest area, 18% of the study watershed, and the PURE Corridor covers less area, 14%. The two strategies have about equal success in predicting locations of Irene landslides and floodplain deposits (> 90%). However, the PURE Corridor covers much more of the drainage network up tributaries, meaning the average PURE Corridor width is smaller than the average ARA width. By design, the PURE Corridor covers 100% of the mapped streams, but the ARA covers only 55%. As a result, important river corridor process are not considered along 45% of the waterways, especially in the headwaters. We do not have specific field observations of each and every river corridor process for loworder streams as we do for Irene landslides and floodplain deposits along the larger White River. Yet the river corridor processes depicted in Figure 1 occur in the stream reaches across the entire NHD dataset, even if they may occur at small enough scales in the smallest streams that they may not present hazards to human infrastructure. These results suggest that the PURE Corridor captures river corridor processes not only laterally from channels but also longitudinally up most, if not all, of the stream networks. The Vermont Corridor and FEMA flood area cover the least area, 4% and 3% of the study watershed, respectively (Table 2), but they also do not cover all of the channels. In this region, these two strategies are valuable tools in that they do not over-predict areas susceptible to floodplain deposition, even though they under-predict areas susceptible to landslides. These two strategies may also under-predict other areas susceptible to other river corridor processes because they do not cover all of the reaches in the drainage network. The study area has 1,452 km of rivers and streams, based on the NHD dataset. The NHD dataset is imperfect, but one of the most reliable for determining the centerlines of the drainage network. The Vermont Corridor covers 820 km (56%) of the channels, indicating that river corridor processes are not considered for almost half of the channels. This percentage will likely increase as Vermont s Small Stream dataset is further reviewed and revised. In contrast, the FEMA floodplain is delineated along merely 251 km (17%) of the drainage network, showing again how the FEMA dataset is not sufficient by itself to capture the suite of processes in river corridors. The Massachusetts River Protection Act is a relatively blunt tool that applies a 200 ft buffer to the channel edge. The advantage is that it is easy to apply and simple to map. This results in a relatively large total area 18% of the total land area in the study area watershed. In the study area, it performs well in predicting areas susceptible to landslides and in covering at least some area along the entire drainage network. However it is notably less effective than the other strategies in predicting areas susceptible to flooding and other floodplain processes. It overlaps with 96 % of the mapped landslide areas but only 53 % of the floodplain deposit areas.

24 Table 2. River corridor delineations compared to field mapped observations of geomorphic change after Hurricane Irene in Vermont s White River, 2011 Area (km 2 ) (%) Total deposit area (in Upper White Study Area) % Deposits in PURE corridor % Deposits in VT Corridor % Deposits in ARA corridor % Deposits in FEMA flood area % Deposits in Massachusetts River Protection area % Total landslide area (in Upper White Study Area) % Landslide area in PURE corridor % Landslide area in VT corridor % Landslide area in ARA corridor % Landslide area in FEMA flood zone % Lanslides in Massachusetts River Protection area % Total area of Region of analysis (Upper White Study Area) PURE Corridor Area % VT Corridor Area % ARA (TNC) Corridor Area % FEMA Q3 Flood Area % MASS Massachusetts River Protection area % Length of Drainage Network Study area PURE Corridor % VT Corridor % ARA (TNC) Corridor % FEMA Q3 Flood % MASS Massachusetts River Protection area % Area of PURE process units in study area % Flood process unit % Wetland process unit % Steep process unit 4.2 3% Migration process unit % Riparian process unit % Total deposit area (in Upper White Study Area) % Deposits in PURE corridor % 4. Discussion The foundation of the process units approach is the notion that a suite of physical and biological processes interact and alter the landscape in river corridors. We strive to incorporate those processes as in a single, unified (yet modular) corridor. The aim is to use relatively simple, transparent, and physicallyand science-based mapping rules in order to develop a method of delineating a local-scale and/or regional-scale river corridor that is comprehensive, reproducible, and seamless across the North Atlantic

25 region. In verifying our mapping rules and the overall approach, the best test is to examine how well the delineated corridor matches with physical evidence of these processes. In three case studies, we have taken several angles to analyze the efficacy of the PURE Corridor in various settings, and the initial results are promising. The results show that the PURE Corridor is useful for watershed and reach scale analysis. The Baker River and Connecticut River examples show that the PURE Corridor encompasses all of the historic channel locations of this actively migrating and avulsing river. The White River example show that the PURE Corridor predicts 87% of areas susceptible to floodplain sedimentation and 92% of areas susceptible to landslides adjacent to river channels in an extreme flood which had, in most places where it was calculated in Vermont, a recurrence interval much, much longer than 100 years. Overall, the PURE Corridor incorporates and covers 100% of the mapped channels longitudinally up watersheds. Further testing of the methodology is warranted, including how well the PURE Corridor delineation strategy predicts locations of riparian ecological processes that are not strictly tied to flooding, floodplain sedimentation, river migration, and steep terrain physical processes. Broader testing is also warranted for the emerging strategies to predict the flood prone areas uniformly over large regions (Wing et al., 2018)(Bent et al., 2015; Clubb et al., 2017; De Roo et al., 2000; Durand et al., 2010). As improved mapping rules and data sources are developed, they can be incorporated into each respective module of the process units approach. In the PURE method we explicitly include two new processes that, to our knowledge, have not been included in any other delineation methodology: a Wetland Processes Unit and Steep Terrain Processes Unit. We feel that these are important additions, because of the well-established benefits of channelproximal wetlands, the rapid delivery of materials to rivers from steep areas, and the hazards of presented by landslides in steep terrain. In addition, the Steep Terrain Processes unit captures the very near-channel areas adjacent to incised rivers, which often fall outside of mapped inundation areas but can be extremely vulnerable to the erosive power of fast moving floodwaters. The PURE method considers the effects of changing climate on the landscape because the corridors depict areas where increased flooding and surface runoff are most likely to accelerate the transfer of material between river channels, floodplains, and the broader landscape. Similarly, the PURE Corridor helps predict areas along river networks most susceptible to increased hazards, including landslides, river migration, and floodplain sedimentation. With changing climate, rivers in the NALCC region are predicted to have more frequent flooding and associated geomorphic change (Huang et al., 2017). Current predictions of climate change do not yet warrant changes in the width of river corridors as we map them. This is due in part to the uncertainty in how climate change predictions will bear out in river systems. For example, the NALCC region is predicted to have continued trends of increased precipitation, increased rainfall compared to snowfall, increased individual storm intensity and increased variability in precipitation that could lead to strengthened wet periods as well as longer dry and/or drought intervals (Parry et al., 2007; Stocker et al., 2013). However, it is still uncertain if changing precipitation will increase both large and small floods, and by how much. The uncertainty is coupled with the existing uncertainty in modeling the extent of flood-prone areas. At present, refining river corridors based on existing climate change information is a poor use of limited budgets, especially since the PURE Corridor already depicts areas most susceptible to climate changes.

26 Because our initial focus was specifically on river processes and how best to map each of them, our process units approach is not influenced by the practical or political feasibility of legislating activities within the river corridor. Eventually, there may be a need or desire to refine these process units to explicitly legislate a compromise between conservation of the maximum practical extent of land surrounding a river (which may look a lot like the PURE Corridor presented in our examples) and something more modest, that allows for more human land uses such as infrastructure and building. Nevertheless, we sought regular feedback on the PURE River Corridor method from the wide array of stakeholders through the RiverSmart FGM Task Force to help ensure that the outcome is useful now and in the future. At present, the PURE River Corridor method can be applied for a wide array of planning and management purposes to avoid natural hazards, allow natural river processes to occur unimpeded, and to protect and conserve fish, wildlife, and the environmentally sensitive riparian habitats they occupy. 5. Acknowledgements This work was supported by the North Atlantic Landscape Conservation Cooperative (WMI), the National Science Foundation, and USDA NIFA Grant # References Allen, P. A., 2008, From landscapes into geological history: Nature, v. 451, no. 7176, p Bales, J., and Wagner, C., 2009, Sources of uncertainty in flood inundation maps: Journal of Flood Risk Management, v. 2, no. 2, p Bent, G. C., Lombard, P. J., and Dudley, R. W., 2015, Flood-Inundation Maps for the North River in Colrain, Charlemont, and Shelburne, Massachusetts, From the Confluence of the East and West Branch North Rivers to the Deerfield River: US Geological Survey, Bieger, K., Rathjens, H., Allen, P. M., and Arnold, J. G., 2015, Development and evaluation of bankfull hydraulic geometry relationships for the physiographic regions of the United States: JAWRA Journal of the American Water Resources Association, v. 51, no. 3, p Bierman, P. R., Montgomery, D. R., University of Vermont., and University of Washington., 2014, Key concepts in geomorphology, New York, NY, W.H. Freeman and Company Publishers : A Macmillan Higher Education Company, xiv, 494, 422 pages p.: Biron, P. M., Buffin-Belanger, T., Larocque, M., Chone, G., Cloutier, C. A., Ouellet, M. A., Demers, S., Olsen, T., Desjarlais, C., and Eyquem, J., 2014, Freedom Space for Rivers: A Sustainable Management Approach to Enhance River Resilience: Environmental Management, v. 54, no. 5, p Bracken, L. J., and Croke, J., 2007, The concept of hydrological connectivity and its contribution to understanding runoff-dominated geomorphic systems: Hydrological Processes, v. 21, no. 13, p Buffin-Belanger, T., Biron, P. M., Larocque, M., Demers, S., Olsen, T., Chone, G., Ouellet, M. A., Cloutier, C. A., Desjarlais, C., and Eyquem, J., 2015, Freedom space for rivers: An economically viable river management concept in a changing climate: Geomorphology, v. 251, p Buraas, E. M., Renshaw, C. E., Magilligan, F. J., and Dade, W. B., 2014, Impact of reach geometry on stream channel sensitivity to extreme floods: Earth Surface Processes and Landforms, v. 39, no. 13, p

27 Castelle, A. J., Johnson, A., and Conolly, C., 1994, Wetland and stream buffer size requirements a review: Journal of Environmental Quality, v. 23, no. 5, p Clubb, F. J., Mudd, S. M., Milodowski, D. T., Valters, D. A., Slater, L. J., Hurst, M. D., and Limaye, A. B., 2017, Geomorphometric delineation of floodplains and terraces from objectively defined topographic thresholds: Earth Surface Dynamics, v. 5, no. 3, p Cowardin, L. M., Carter, V., Golet, F. C., and LaRoe, E. T., 1979, Classification of wetlands and deepwater habitats of the United States: US Department of the Interior, US Fish and Wildlife Service. Croke, J., Fryirs, K., and Thompson, C., 2013, Channel-floodplain connectivity during an extreme flood event: implications for sediment erosion, deposition, and delivery: Earth Surface Processes and Landforms, v. 38, no. 12, p De Roo, A., Wesseling, C., and Van Deursen, W., 2000, Physically based river basin modelling within a GIS: the LISFLOOD model: Hydrological Processes, v. 14, no , p Dethier, E., Magilligan, F. J., Renshaw, C. E., and Nislow, K. H., 2016, The role of chronic and episodic disturbances on channel-hillslope coupling: the persistence and legacy of extreme floods: Earth Surface Processes and Landforms, v. 41, no. 10, p Durand, M., Fu, L.-L., Lettenmaier, D. P., Alsdorf, D. E., Rodriguez, E., and Esteban-Fernandez, D., 2010, The surface water and ocean topography mission: Observing terrestrial surface water and oceanic submesoscale eddies: Proceedings of the IEEE, v. 98, no. 5, p Fischer, R. A., and Fischenich, J. C., 2000, Design recommendations for riparian corridors and vegetated buffer strips: Army engineer waterways experiment station vicksburg ms engineer research and development center. Gartner, J. D., 2015, Connectivity of sediment in river systems physical controls and predictions of types, timescales and magnitudes [Ph.D. Dissertation: Dartmouth College, 200 p. Gartner, J. D., Dade, W. B., Renshaw, C. E., Magilligan, F. J., and Buraas, E. M., 2015, Gradients in stream power influence lateral and downstream sediment flux in floods: Geology, v. 43, no. 11, p Gartner, J. D., Mersel, M. K., Lefebvre, L. E., and Lichvar, R. W., 2016, The Benefits and Limitations of Hydraulic Modeling for Ordinary High Water Mark Delineation: US Army Engineer Research and Development Center (ERDC) Hanover United States. Gesch, D., Oimoen, M., Greenlee, S., Nelson, C., Steuck, M., and Tyler, D., 2002, The national elevation dataset: Photogrammetric Engineering and Remote Sensing, v. 68, no. 1, p Hickin, E. J., and Nanson, G. C., 1984, Lateral migration rates of river bends: Journal of Hydraulic Engineering, v. 110, no. 11, p Huang, H., Winter, J. M., Osterberg, E. C., Horton, R. M., and Beckage, B., 2017, Total and extreme precipitation changes over the Northeastern United States: Journal of Hydrometeorology, v. 18, no. 6, p Hupp, C. R., and Osterkamp, W., 1996, Riparian vegetation and fluvial geomorphic processes: Geomorphology, v. 14, no. 4, p Kline, M., and Cahoon, B., 2010, Protecting river corridors in Vermont: JAWRA Journal of the American Water Resources Association, v. 46, no. 2, p Kline, M., and Dolan, K., 2008, River corridor protection guide: Fluvial geomorphic-based methodology to reduce flood hazards and protect water quality: Vermont Agency of Natural Resources. Montpelier, VT. Knighton, D., 2014, Fluvial forms and processes: a new perspective, Routledge. Landis, J. D., Hamm, N. T., Renshaw, C. E., Dade, W. B., Magilligan, F. J., and Gartner, J. D., 2012, Surficial redistribution of fallout 131 iodine in a small temperate catchment: Proceedings of the National Academy of Sciences of the United States of America, v. 109, no. 11, p

28 Larsen, E. W., 2007, Sacramento River Ecological Flows Study: Meander Migration Modeling Final Report: Prepared for The Nature Conservancy, Chico, CA by Eric W. Larsen, Davis, CA. Lee, P., Smyth, C., and Boutin, S., 2004, Quantitative review of riparian buffer width guidelines from Canada and the United States: Journal of Environmental Management, v. 70, no. 2, p Lichvar, R. W., and McColley, S. M., 2008, A Field Guide to the Identification of the Ordinary High Water Mark (OHWM) in the Arid West Region of the Western United States: A Delineation Manual: Engineer Research and Development Center Hanover NH Cold Regions Research and Engineering Lab. Lombard, P. J., and Bent, G. C., 2015, Flood-Inundation Maps for the Deerfield River, Franklin County, Massachusetts, From the Confluence With the Cold River Tributary to the Connecticut River: U.S. Geological Survey Scientific Investigations Report , 22 p., appendixes. Magilligan, F. J., Buraas, E., and Renshaw, C., 2015, The efficacy of stream power and flow duration on geomorphic responses to catastrophic flooding: Geomorphology, v. 228, p McCormick, S. D., Hansen, L. P., Quinn, T. P., and Saunders, R. L., 1998, Movement, migration, and smolting of Atlantic salmon (Salmo salar): Canadian Journal of Fisheries and Aquatic Sciences, v. 55, p McKay, L., Bondelid, T., Dewald, T., Johnston, J., Moore, R., and Rea, A., 2012, NHDPlus Version 2: user guide: National Operational Hydrologic Remote Sensing Center, Washington, DC. Micheli, E., Kirchner, J., and Larsen, E., 2004, Quantifying the effect of riparian forest versus agricultural vegetation on river meander migration rates, Central Sacramento River, California, USA: River Research and Applications, v. 20, no. 5, p Milly, P. C. D., Betancourt, J., Falkenmark, M., Hirsch, R. M., Kundzewicz, Z. W., Lettenmaier, D. P., and Stouffer, R. J., 2008, Climate change - Stationarity is dead: Whither water management?: Science, v. 319, no. 5863, p Milly, P. C. D., Wetherald, R. T., Dunne, K. A., and Delworth, T. L., 2002, Increasing risk of great floods in a changing climate: Nature, v. 415, no. 6871, p Naiman, R. J., and Decamps, H., 1997, The ecology of interfaces: riparian zones: Annual review of Ecology and Systematics, v. 28, no. 1, p Naiman, R. J., Decamps, H., and Pollock, M., 1993, The role of riparian corridors in maintaining regional biodiversity: Ecological applications, v. 3, no. 2, p Novitski, R., Smith, R. D., and Fretwell, J. D., 1996, Wetland functions, values, and assessment: National Summary on Wetland Resources. USGS Water Supply Paper, v. 2425, p Ollero, A., 2010, Channel changes and floodplain management in the meandering middle Ebro River, Spain: Geomorphology, v. 117, no. 3-4, p Parry, M., Parry, M. L., Canziani, O., Palutikof, J., Van der Linden, P., and Hanson, C., 2007, Climate change 2007-impacts, adaptation and vulnerability: Working group II contribution to the fourth assessment report of the IPCC, Cambridge University Press. Pasquale, N., Perona, P., Schneider, P., Shrestha, J., Wombacher, A., and Burlando, P., 2011, Modern comprehensive approach to monitor the morphodynamic evolution of a restored river corridor: Hydrology and Earth System Sciences, v. 15, no. 4, p Pickett, S. T., and White, P. S., 1985, The ecology of natural disturbance and patch dynamics, Orlando, Fla., Academic Press. Piegay, H., Darby, S. E., Mosselman, E., and Surian, N., 2005, A review of techniques available for delimiting the erodible river corridor: A sustainable approach to managing bank erosion: River Research and Applications, v. 21, no. 7, p Pitlick, J., 1997, A regional perspective of the hydrology of the 1993 Mississippi River basin floods: Annals of the Association of American Geographers, v. 87, no. 1, p

29 Pringle, C. M., 2001, Hydrologic connectivity and the management of biological reserves: A global perspective: Ecological Applications, v. 11, no. 4, p Rapp, C., and Abbe, T., 2003a, A Framework or Delineating Channel Migration Zones, Washington State Department of Ecology Report , WS. Rapp, C. F., and Abbe, T. B., 2003b, A framework for delineating channel migration zones. Rawlins, M., Bradley, R., and Diaz, H., 2012, Assessment of regional climate model simulation estimates over the northeast United States: Journal of Geophysical Research: Atmospheres, v. 117, no. D23. Ridge, J. C., and Larsen, F. D., 1990, Re-evaluation of Antevs' New England varve chronology and new radiocarbon dates of sediments from glacial Lake Hitchcock: Geological Society of America Bulletin, v. 102, no. 7, p Rose, S., and Peters, N. E., 2001, Effects of urbanization on streamflow in the Atlanta area (Georgia, USA): a comparative hydrological approach: Hydrological Processes, v. 15, no. 8, p Samela, C., Albano, R., Sole, A., and Manfreda, S., 2018, A GIS tool for cost-effective delineation of floodprone areas: Computers, Environment and Urban Systems, v. 70, p Schwenk, J., Khandelwal, A., Fratkin, M., Kumar, V., and Foufoula-Georgiou, E., 2017, High spatiotemporal resolution of river planform dynamics from Landsat: The RivMAP toolbox and results from the Ucayali River: Earth and Space Science, v. 4, no. 2, p Smith, M. P., Schiff, R., Olivero, A., and MacBroom, J., 2008, The active river area: A conservation framework for protecting rivers and streams: The Nature Conservancy, Boston, MA. Stanford, J. A., and Ward, J. V., 1993, An Ecosystem Perspective of Alluvial Rivers - Connectivity and the Hyporheic Corridor: Journal of the North American Benthological Society, v. 12, no. 1, p Stocker, T. F., Qin, D., Plattner, G., Tignor, M., Allen, S., Boschung, J., Nauels, A., Xia, Y., Bex, V., and Midgley, P., 2013, Climate change 2013: the physical science basis. Intergovernmental panel on climate change, working group I contribution to the IPCC fifth assessment report (AR5): New York. Strayer, D. L., 2008, Freshwater mussel ecology : a multifactor approach to distribution and abundance, Berkeley, University of California Press, Freshwater ecology series, v. 1, ix, 204 p. p.: USFWS, 1997, A system for mapping riparian areas in the western United States: US Fish and Wildlife Service, National Wetlands Inventory, Denver. Ward, P. J., Jongman, B., Aerts, J. C. J. H., Bates, P. D., Botzen, W. J. W., Loaiza, A. D., Hallegatte, S., Kind, J. M., Kwadijk, J., Scussolini, P., and Winsemius, H. C., 2017, A global framework for future costs and benefits of river-flood protection in urban areas: Nature Climate Change, v. 7, no. 9, p Warner, B. P., Gartner, J. D., and Hatch, C. E., 2018, Fluvial geomorphic assessment and river corridor mapping as flood risk management tools in Massachusetts, USA: Journal of Flood Risk Management, v. 11, p. S1100-S1104. Watson, K. B., Ricketts, T., Galford, G., Polasky, S., and O'Niel-Dunne, J., 2016, Quantifying flood mitigation services: The economic value of Otter Creek wetlands and floodplains to Middlebury, VT: Ecological Economics, v. 130, p Wilen, B. O., and Bates, M., 1995, The US fish and wildlife service s national wetlands inventory project, Classification and inventory of the world s wetlands, Springer, p Wing, O. E., Bates, P. D., Smith, A. M., Sampson, C. C., Johnson, K. A., Fargione, J., and Morefield, P., 2018, Estimates of present and future flood risk in the conterminous United States: Environmental Research Letters, v. 13, no. 3, p Wohl, E., 2017, The significance of small streams: Frontiers of Earth Science, v. 11, no. 3, p Wohl, E., and Beckman, N. D., 2014, Leaky rivers: Implications of the loss of longitudinal fluvial disconnectivity in headwater streams: Geomorphology, v. 205, p

30 Yellen, B., Woodruff, J. D., Cook, T. L., and Newton, R. M., 2016, Historically unprecedented erosion from Tropical Storm Irene due to high antecedent precipitation: Earth Surface Processes and Landforms, v. 41, no. 5, p

31 Figure 1. Theoretical landscape (a) showing cross-sectional views of river corridor processes in (b) a narrow, constrained bedrock valley with little alluvium and (c) a broad, self-formed valley dominated by alluvium. A plan view of the river at this location is also shown. River processes are identified by number in both c and d, and correspond to the following: (1) Flooding, (2) River migration, (3) Wetland processes including flood water retention and ecosystem services, (4) Overbank deposition, (5) Point bar deposition, (6) Bank erosion, (7) Avulsions, (8) In-channel erosion (incision) and deposition (aggradation), and (9) Landslides, bank failures and debris flows.

32 a Figure 2a. This photograph by Andrew Drummond on October 30, 2017 shows flooding of the Saco River in Maine, illustrating the inundation associated with the Flood Processes Unit.

33 b Figure 2b. This photograph by Christine Hatch on May 19, 2016 shows land sliding and trees collapsing into the Chickley River in Massachusetts, illustrating bank failures associated with the Landslide and Steep Terrain Processes Unit.

34 c Figure 2c. This photograph by NYS DEC on August 31, 2011 shows a historic wetland full of floodwaters after Hurricane Irene in the Adirondack Mountains, New York. Wetland areas connected to river systems are captured by the Wetland Processes Unit.

35 Figure 2d. The Channel Migration Processes Unit represents the meander migration rates and formation of off-channel wet areas such as oxbows over time, illustrated here by Google Earth imagery along the Connecticut River in New Hampshire, near (a) Haverhill, (b) South Lancaster, and (c) Northumberland (2009), respectively.

36 e Figure 2e. American Forests is protecting riparian areas along small streams in Vermont near Lake Champlain, as seen in this photo by Eric Sprague. Other important riparian areas not included in the other units are captured in the Riparian Ecologic Processes Unit, which can be configured for specific conservation goals.

37 Figure 3. Available Federal Emergency Management Agency (FEMA) Flood Hazard Layer (FHL) maps in the Northeastern region of the U.S. as see in Google earth. Note, there are large areas with no coverage.

38 Figure 4. Schematic representation of how individual modular process units are mapped, overlain and summed to form the total Process Units River Evaluation (PURE) River Corridor.

39 Figure 5. The upper Connecticut River test location. (a) USGS Topographic Map near Haverhill, New Hampshire.

40 Figure 5. The upper Connecticut River test location. (b) Google earth satellite imagery.

41 Figure 5. The upper Connecticut River test location. (c) GIS Layer showing USGS Digital Elevation Model (DEM) hillshade basemap overlain by the National Hydrolography Dataset (NHD) channel flowlines (dark blue) and, for wide channels, channel area (light blue).

42 Figure 5. The upper Connecticut River test location. (d) GIS Layer showing USGS Digital Elevation Model (DEM) hillshade basemap overlain by Process Units River Evaluation (PURE) river corridor layers: Flood Processes Unit (light blue) and Landslide and Steep Terrain Process Unit (orange).

43 Figure 5. The upper Connecticut River test location. (e) GIS Layer showing USGS Digital Elevation Model (DEM) hillshade basemap overlain by Process Units River Evaluation (PURE) river corridor layer: Flood Processes Unit (light blue) and Wetland Process Unit (dark blue).

44 Figure 5. The upper Connecticut River test location. (f) GIS Layer showing USGS Digital Elevation Model (DEM) hillshade basemap overlain by Process Units River Evaluation (PURE) river corridor layer: River Migration Process Unit (light yellow).

45 Figure 5. The upper Connecticut River test location. (g) GIS Layer showing USGS Digital Elevation Model (DEM) hillshade basemap overlain by Process Units River Evaluation (PURE) river corridor layer: Riparian Process Unit (olive green).

46 Figure 5. The upper Connecticut River test location. (h) GIS Layer showing USGS Digital Elevation Model (DEM) hillshade basemap overlain by the total Process Units River Evaluation (PURE) river corridor layer (pink).

47 Figure 5. The upper Connecticut River test location. (i) GIS Layer showing USGS Digital Elevation Model (DEM) hillshade basemap overlain by the total Process Units River Evaluation (PURE) river corridor layer (pink) and the Vermont River Corridor* (light orange). Note: The Vermont River Corridor does not map the valley bottom of the Connecticut River, nor the New Hampshire side of the river.

48 Figure 5. The upper Connecticut River test location. (j) GIS Layer showing USGS Digital Elevation Model (DEM) hillshade basemap overlain by the total Process Units River Evaluation (PURE) river corridor layer (pink) and The Nature Conservancy s (TNC) Active River Area (ARA; green).

49 Figure 5. The upper Connecticut River test location. (k) GIS Layer showing USGS Digital Elevation Model (DEM) hillshade basemap overlain by the total Process Units River Evaluation (PURE) river corridor layer (light pink) and the Massachusetts Rivers Protection Act maximum buffer (200-feet; dark pink).

50 Figure 6. The Baker River test location, near Warren, Vermont. Google earth satellite imagery overlain by historical channel positions from 1993 (orange), 2003 (pink), 2011 (yellow), 2013 (blue) and 2015 (green). The Total Process Units River Evaluation (PURE) river corridor layer is outlined in white, and the PURE Flood Process Unit alone is outlined in blue.

51 Figure 7. The White River test location near Rochester, Vermont. (a) Google earth satellite imagery

52 b Figure 7. The White River test location near Rochester, Vermont. (b) Google terrain map.

53 c Figure 7. The White River test location near Rochester, Vermont. GIS Layer showing USGS Digital Elevation Model (DEM) hillshade basemap overlain by the field-mapped extent of features that occurred as a result of Hurricane Irene in August 2011, including: erosion from landslide scarps (dark red) and sediment deposition (yellow). Underneath these features, for comparison, is (c) Total Process Units River Evaluation (PURE) river corridor layer (pink).